Agentic AEO is live on the Shopify App Store

Measurement

The AI Commerce Index, taken apart.

One number between 0 and 100, with every layer under it left visible.

Run a free auditHow the score works
Representative demo · Threshold Footwear
Representative demo · Threshold Footwear

One question set holds still.

The other moves. Reputation is controlled; Discovery is your board, and meant to change.

Reputation battery

Fixed set · never turns over

The questions
Twelve, fixed, identical for every merchant. Only the brand name and category change.
What it answers
What the assistants say about you: legitimacy, safety, reviews, returns, shipping, warranty and six more.
Firing order
Round-robin, stalest question next. A full cycle every 72 hours, on every plan.
What the engine sees
Your brand name, never your domain. It has to recognise you the way a customer would.
Turnover
None. All twelve fire the day you install, then rotate for as long as you are on Burnish.
Feeds Representation

Discovery board

Open set · rotates continuously

The questions
Open-ended questions a shopper would type — up to 50 on Presence and Authority, 100 on Omnipresence.
What it answers
Whether you are in the answer when someone asks for products like yours.
Firing order
Weighted by what a question is worth and how long since it was read. One firing every 120, 90 or 35 minutes by plan.
What the engine sees
The question as a shopper would put it, and nothing identifying the store asking.
Turnover
A question sits on the board at least 14 days before it can rotate out.
Feeds Presence

Two halves, weighted.

What your catalogue hands an engine, and what the engines said back.

Index = Readiness × 0.4 + Visibility × 0.6

Readiness × 0.4

Eight components, every one a field you control and Burnish can close.

Visibility × 0.6

What five assistants said when asked, scored over a rolling 21 days.

Before the first reading

Readiness alone. It never averages in a number it has not taken.

The inner blend is not published

Representation and Presence make up visibility. That split is still being calibrated, so Burnish does not print it.

When Burnish refuses.

Four rules that stop a number appearing before it means anything.

  1. Under ten samples, no band

    The screen says so, rather than drawing confidence around three readings.

  2. Twenty-one days, not yesterday

    Both halves score over a rolling window. One unusual answer cannot swing it.

  3. Confidence travels with the score

    The interval is carried into the index rather than dropped at the last step.

  4. Rubric versions are not silently compared

    When the rubric changes shape, earlier scores stop being comparable. Burnish says so.

Representative demo · Threshold Footwear
Representative demo · Threshold Footwear
The app, refusing — Reputation · Threshold Footwear fixture

Named and right are different failures.

A store can be recommended constantly and described wrongly. Separate numbers, separate work.

Reputation

How many engines recognise the brand, and how well they speak about it. A brand none recognises scores zero.

Discovery

Named outright, or only landing near. Proximity alone is capped at 45.

Representative demo · Threshold Footwear
Representative demo · Threshold Footwear

Agentic AEO runs on Burnish.

Free AI auditFree AI audit